Handover ratio

Confusion matrix


举个例子 现在有两类 橘子 5个 苹果 5个
准确率就是 正确检测是橘子/(正确检测是橘子+错误检测是橘子)   判断的是,橘子你检测的准不准
召回率就是 我这五个橘子结果你只检测出了四个橘子 判断的是你检测的全不全 

IOU

吻合程度:预测框和真实框的交集/并集

Corresponds to accuracy rate and recall rate

AP (Average Precision)

The average accuracy of a certain category under different recall rates

MAP Average AP of multiple categories

Guess you like

Origin blog.csdn.net/ahelloyou/article/details/112971253